Writing

Papers and notes

8 posts on what I learned building these systems, and 4 published papers. Every post says what broke and what the number was afterward.

Posts

  • Post

    Do I have a guarantee, or am I guessing?

    Convergence analysis taught me to ask one question of any algorithm. That question shaped an agentic supply-chain system built to decide when not to trust its own answers.

    • Agentic AI
    • Convergence analysis
    • AI safety
  • Post

    A hand-tuned prompt scored 91%. A compiled one scored 99%.

    Route a supply-chain question wrong and an inventory query comes back with a vehicle-routing answer. I replaced the hand-written classifier prompt with a DSPy-compiled one, then spent longer on the measurement than on the compile.

    • DSPy
    • Evaluation
    • Agentic AI
  • Post

    Same question, different numbers, wrong answer

    "Allocate 400 units" and "allocate 1,000 units" embed almost identically. A cache that matches on similarity alone will serve one manager another manager's numbers.

    • Agentic AI
    • Semantic caching
    • Production ML
  • Post

    Why we proved convergence on the simplest network there is

    The linear layer is the one component every neural network shares. Its loss is still nonconvex, and training it is still NP-complete. That is why it is the right place to look for a guarantee.

    • Deep learning
    • Optimization
    • Convergence analysis
  • Post

    I defended my PhD: convergence analysis of training deep linear neural networks

    The learning rate depends on constants existing theory can't compute before training, so they get chosen by trial and error. I closed part of that gap, and I'm now applying the same rigor to production AI.

    • Deep learning
    • Optimization theory
    • Convergence analysis
  • Post

    The copilot keeps answering when the language model goes dark

    A free tier capped at 50 API calls a day forced every design decision in this system. The constraint is gone now. The architecture it produced stayed, because it earned its place.

    • Agentic AI
    • System design
    • Production ML
  • Post

    The most reliable part of the copilot is the part the LLM never touches

    Ask a language model to plan a logistics decision and you get the most statistically likely answer, not the cheapest one. So I kept the model out of the math and spent the effort on the boundary between them.

    • Agentic AI
    • Operations research
    • Neuro-symbolic AI
  • Post

    $10,000 is where the agent stops and a human decides

    Every approval workflow needs a number. Below it the agent executes, above it a person signs off. The bug that taught me where to put that check only surfaced in integration testing.

    • Agentic AI
    • Human in the loop
    • Enterprise AI

Published papers

Paper

Convergence of gradient descent for learning linear neural networks

Springer Nature: Advances in Continuous and Discrete Models, 2024

Gradient descent converges to a global minimum for deep linear networks under an explicit step-size bound that stops shrinking as the network deepens. The paper carries earlier gradient flow results into the discrete setting, and drops the deficiency-margin and boundedness assumptions the earlier work needed.